The matching engine is only as good as the data behind it. You can build the most sophisticated algorithm in the world, but if your venue database is garbage, you’re matching people to garbage.
So we needed to build the most comprehensive database of Manhattan venues that exists. Not the biggest. The most comprehensive. Every bar, restaurant, cafe, lounge, club, music venue, comedy club, rooftop, speakeasy, food hall, and activity space. With data that goes way beyond name and address.
We started with public data sources. Business listings. Open datasets. APIs. Anything we could legally scrape or pull. The raw data was a mess. Duplicate entries. Closed businesses. Wrong addresses. Venues listed under corporate parent names instead of the name on the door.
The deduplication alone took days. "Joe’s Pizza" in the West Village and "Joe’s Pizza Inc." in the West Village and "JOES PIZZA" in the West Village — that’s one place, not three. Multiply that by 25,000 and you’ll understand why we stopped sleeping.
Then came the enrichment.
A name and address isn’t enough. We need mood data. What’s the vibe? Is it loud or quiet? Is it date-night material or group-hangout material? Is it trendy or classic? What’s the crowd like? What’s the best time to go?
Some of this we could infer from categories and descriptions. A "craft cocktail lounge" in the Lower East Side with 30 seats gets tagged differently than a "sports bar and grill" in Midtown East with 200 seats. But a lot of it required manual work. Going through venues one by one, cross-referencing photos, descriptions, reviews, and our own knowledge of the city.
We wrote scripts to help. Natural language processing on descriptions to extract vibe keywords. Price estimation from menu data. Neighborhood assignment from geocoordinates mapped against custom boundary polygons (more on that in a future post).
The result: 25,000+ venues, each with structured data including:
Name, address, coordinates. Phone, website, hours. Category tags (bar, restaurant, cafe, etc.). Neighborhood assignment across all 43 Manhattan neighborhoods. Mood and vibe tags. Price tier. Capacity estimate. Best-for scenarios. Insider tips where available.
Is it perfect? No. Some venues have thin data. Some hours are wrong. Some places closed last week and we don’t know yet. This is a living database — it’ll get better every day.
But it’s real. It’s deep. And it’s mine.
Nobody else has this. The big platforms have listings, but they don’t have mood data. They can tell you a bar has 4.2 stars. They can’t tell you it’s perfect for a first date on a Tuesday when you want something intimate and under $40.
That’s the gap. That’s what 25,000 venue profiles with mood data fills.
Next up: building the engine that actually matches you to these venues based on how you feel right now.
— The Moodap™ Team